Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Task Analysis”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,207 records · Page 67
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

Progress Towards NQA-1 for Cardinal in FY25

Cardinal is a wrapping of the GPU-oriented spectral element Computational Fluid Dynamics (CFD) code NekRS and the Monte Carlo particle transport code OpenMC within the Multiphysics Object-Oriented Simulation Environment (MOOSE). Cardinal provides high-resolution thermal-hydraulics and/or radiation transport feedback to MOOSE multiphysics simulations. Multiphysics feedback is implemented in a geometry-agnostic manner which eliminates the need for rigid one-to-one mappings. A generic data transfer implementation also allows NekRS and OpenMC to couple to any MOOSE application, enabling a broad set of multiphysics capabilities. Cardinal simulations can also leverage combinations of MPI, OpenMP, and GPU resources. Cardinal continuous development and improvement efforts have led to the software being considered as a high-fidelity design and licensing tool for key areas of nuclear reactor relevant physics, including neutron transport, fluid flow, heat transfer, and mechanical processes. The fast development and expansion of the software from a pure R&D framework towards its application in the nuclear industry and regulation require a focus on developing, enhancing,and maintaining Cardinal’s software quality through strict adherence to a Software Quality Assurance (SQA) framework and SQA program. To facilitate compliance with SQA standards, the Cardinal SQA Program was initiated during Fiscal Year 2023 (FY23). During the development of the Cardinal SQA Program, multiple gaps have been identified. These gaps are primarily related to model verification and code pedigree as they relate to the use of Cardinal as an analysis tool. These gaps were captured in a report published in 2023. A second report highlighted the progress made during Fiscal Year 2024 (FY24) and described Argonne’s effort to document and integrate software verification within Cardinal’s software development process. This report documents the progress made towards NQA-1 for Cardinal in the Fiscal Year 2025 (FY25). All cases in the expanded Continuous Integration (CI) suite of NekRS are included in this report which test the solvers and modules available in NekRS exhaustively. The NekRS tests are integrated with the Cardinal CI suite and made available in publicly accessible Github documentation. Following the CI practice permits integrating of source code changes frequently and ensuring that the integrated codebase clears the verification testing for the software. Also in this report is a brief overview of the development of the Cardinal Software Quality Assurance Plan (SQAP) that was done in FY25, though it should be noted that the rest of the documentation for the SQA program needs to be developed in a future step of this task.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2024

Solving the Orszag–Tang vortex magnetohydrodynamics problem with physics-constrained convolutional neural networks

We study the 2D Orszag–Tang vortex magnetohydrodynamics (MHD) problem through the use of physics-constrained convolutional neural networks (PCNNs) for forecasting the density, ρ, and the magnetic field, B, as well as the prediction of B given the velocity field v of the fluid. In addition to translation equivariance from the convolutional architecture, other physics constraints were embedded: absence of magnetic monopoles, non-negativity of ρ, use of only relevant variables, and the periodic boundary conditions of the problem. The use of only relevant variables and the hard constraint of non-negative ρ were found to facilitate learning greatly. The divergenceless condition ∇·B=0 was implemented as a hard constraint up to machine precision through the use of a magnetic potential to define B=∇×A. Residual networks and data augmentation were also used to improve performance. This allowed for some of the residual models to function as surrogate models and provide reasonably accurate simulations. For the prediction task, the PCNNs were evaluated against a physics-informed neural network, which had the ideal MHD induction equation as a soft constraint. Several models were able to generate highly accurate fields, which are visually almost indistinguishable and have low mean squared error. Only methods with built-in hard constraints produced physical fields with ∇·B=0. The use of PCNNs for MHD has the potential to produce physically consistent real-time simulations to serve as virtual diagnostics in cases where inferences must be made with limited observables.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2020

Deciphering the PMI Surface Chemistry of Lithium-Based PFCs and its Effects on High Performance Plasmas in NSTX

The primary aims or objectives of the proposed work here focus on the dynamic measurement (Illinois PI Allain) and atomistic-based multi-scale computational atomistic simulations (UTK PI: Wirth) of: 1) lithium coatings with re-deposited hydrogenated carbon surfaces and high-Z substrates, 2) lithium coatings on boronized ATJ graphite surfaces and high-Z substrates and 3) lithium coatings on ATJ graphite and high-Z substrates with variation in temperature. Three main primary tasks are proposed to address each of the primary objectives listed above: 1) dynamic in-situ irradiation with D+, He+ species of lithium coatings with variation in interface substrate morphology (e.g. smooth vs rough, fuzz vs nanostructured); 2) validation and coupling with multi-scale computational simulations connecting the irradiated surface to the plasma edge; 3) extrapolation to long-pulse conditions (e.g. function of flux vs fluence studies). In primary objective #3, temperature-based effects will be studied to evaluate three primary mechanisms: a) D uptake and recycling, b) erosion, and c) surface melting and evaporation as a function of substrate morphology and chemistry.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2021

WOMBAT (Windfarm Operations and Maintenance cost-Benefit Analysis Tool) WISDEM® [SWR-21-68]

The windfarm operations and maintenance cost-benefit analysis tool (WOMBAT) is software to simulate the operations and maintenance phase of either onshore or offshore windfarms. WOMBAT is a medium-fidelity tool that is designed to model turbine, cable, and substation failures at the subassembly level, but is flexible enough that individual component or asset-level modeling can also be performed. WOMBAT enables users to model technological improvements, the use of a wide range of service equipment from drone repairs and remote resets to heavy lift vessels and crawler cranes, to analyze cost trends with varying trade-offs in conjunction with energy production, availability and a growing range of metrics. This library provides a tool to simulate the operation and maintenance phase (O&M) of distributed, land-based, and offshore windfarms using a discrete event simulation framework. WOMBAT is written around the SimPy discrete event simulation framework. Additionally, this is supported using a flexible and modular object-oriented code base, which enables the modeling of arbitrarily large (or small) windfarms with as many or as few failure and maintenance tasks that can be encoded. Please note that this is still heavily under development, so you may find some functionality to be incomplete at the current moment, but rest assured the functionality is expanding. With that said, it would be greatly appreciated for issues or PRs to be submitted for any improvements at all, from fixing typos (guaranteed to be a few) to features to testing. Also available at: https://pypi.org/project/wombat/

Cooperman, Aubryn↗
DOE OSTI2020

Model reduction for hypersonic aerodynamics via conservative LSPG projection and hyper-reduction

High-speed aerospace engineering applications rely heavily on computational fluid dynamics (CFD) models for design and analysis due to the expense and difficulty of flight tests and experiments. This reliance on CFD models necessitates performing accurate and reliable uncertainty quantification (UQ) of the CFD models. However, it is very computationally expensive to run CFD for hypersonic flows due to the fine grid resolution required to capture the strong shocks and large gradients that are typically present. Furthermore, UQ approaches are “many-query” problems requiring many runs with a wide range of input parameters. One way to enable computationally expensive models to be used in such many-query problems is to employ projection-based reduced-order models (ROMs) in lieu of the (high-fidelity) full-order model. In particular, the least-squares Petrov–Galerkin (LSPG) ROM (equipped with hyper-reduction) has demonstrated the ability to significantly reduce simulation costs while retaining high levels of accuracy on a range of problems including subsonic CFD applications. This allows computationally inexpensive LSPG ROM simulations to replace the full-order model simulations in UQ studies, which makes this many-query task tractable, even for large-scale CFD models. This work presents the first application of LSPG to a hypersonic CFD application. In particular, we present results for LSPG ROMs of the HIFiRE-1 in a three-dimensional, turbulent Mach 7.1 flow, showcasing the ability of the ROM to significantly reduce computational costs while maintaining high levels of accuracy in computed quantities of interest.

Fluid Dynamics↗
DOE OSTI2023

Enhancing Monte Carlo Workflows for Nuclear Reactor Analysis with Metamodel-Driven Modeling

Monte Carlo codes are essential components of many reactor physics simulation workflows as high-fidelity continuous-energy neutron transport solvers. Among Monte Carlo radiation transport codes, MCNP is particularly notable due to its diverse simulation capabilities, large user base, and long validation history. Despite being a powerful simulation tool, MCNP provides limited capabilities to allow automated execution, model transformation, or support for user-defined logic and abstractions that limit its compatibility with modern workflows. Here, to better integrate MCNP into a modern scientific workflow, we have developed an intuitive yet full-featured MCNP Application Program Interface (API) in Python, named MCNPy, which provides a specialized set of classes for MCNP input development. Moreover, to guarantee that our reading, writing, and modeling capabilities remain self-consistent (and to render the huge scope of the MCNP API manageable), we have adopted a strategy of model-driven software development in which a generalized model of the MCNP input format has been created. From this generalized model, or “metamodel,” problem-specific implementations such as an engine for input validation or a codebase for programmatic operations may be automatically generated. Since MCNPy primarily acts as a Python front-end to the underlying Java API that directly interfaces with the metamodel, it is intrinsically linked to the metamodel and thus remains maintainable. With MCNPy, users can programmatically read, write, and modify any syntactically valid MCNP input file regardless of its origin. These capabilities allow users to automate complicated tasks like design optimization and model translation for nuclear systems. As examples, this work demonstrates the use of MCNPy to find the critical radius of a plutonium sphere and to translate a 9000+ line MCNP input file into a corresponding OpenMC model.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2023

Process-Oriented Diagnostics: Principles, Practice, Community Development, and Common Standards

Process-oriented diagnostics (PODs) aim to provide feedback for model developers through model analysis based on physical hypotheses. However, the step from a diagnostic based on relationships among variables, even when hypothesis driven, to specific guidance for revising model formulation or parameterizations can be substantial. The POD may provide more information than a purely performance-based metric, but a gap between POD principles and providing actionable information for specific model revisions can remain. Furthermore, in coordinating diagnostics development, there is a trade-off between freedom for the developer, aiming to capture innovation, and near-term utility to the modeling center. Best practices that allow for the former, while conforming to specifications that aid the latter, are important for community diagnostics development that leads to tangible model improvements. Promising directions to close the gap between principles and practice include the interaction of PODs with perturbed physics experiments and with more quantitative process models as well as the inclusion of personnel from modeling centers in diagnostics development groups for immediate feedback during climate model revisions. Examples are provided, along with best-practice recommendations, based on practical experience from the NOAA Model Diagnostics Task Force (MDTF). Common standards for metrics and diagnostics that have arisen from a collaboration between the MDTF and the Department of Energy’s Coordinated Model Evaluation Capability are advocated as a means of uniting community diagnostics efforts.

58 GEOSCIENCES↗
DOE OSTI2023

Application of Electromagnetic Methods for Reservoir Monitoring with Emphasis on Carbon Capture, Utilization, and Storage

The Controlled-Source ElectroMagnetic (CSEM) method provides crucial information about reservoir fluids and their spatial distribution. Carbon dioxide (CO2) storage, enhanced oil recovery (EOR), geothermal exploration, and lithium exploration are ideal applications for the CSEM method. The versatility of CSEM permits its customization to specific reservoir objectives by selecting the appropriate components of a multi-component system. To effectively tailor the CSEM approach, it is essential to determine whether the primary target reservoir is resistive or conductive. This task is relatively straightforward in CO2 monitoring, where the injected fluid is resistive. However, for scenarios involving brine-saturated (water-wet) or oil-wet (carbon capture, utilization, and storage—CCUS) reservoirs, consideration must also be given to conductive reservoir components. The optimization of data acquisition before the survey involves analyzing target parameters and the sensitivity of multi-component CSEM. This optimization process typically includes on-site noise measurements and 3D anisotropic modeling. Based on our experience, subsequent surveys tend to proceed smoothly, yielding robust measurements that align with scientific objectives. Other critical aspects to be considered are using magnetotelluric (MT) measurements to define the overall background resistivities and integrating real-time quality assurance during data acquisition with 3D modeling. This integration allows the fine tuning of acquisition parameters such as acquisition time and necessary repeats. As a result, data can be examined in real-time to assess subsurface information content while the acquisition is ongoing. Consequently, high-quality data sets are usually obtained for subsequent processing and initial interpretation with minimal user intervention. The implementation of sensitivity analysis during the inversion process plays a pivotal role in ensuring that the acquired data accurately respond to the target reservoirs’ expected depth range. To elucidate these concepts, we present an illustrative example from a CO2 storage site in North Dakota, USA, wherein the long-offset transient electromagnetic method (LOTEM), a variation of the CSEM method, and the MT method were utilized. This example showcases how surface measurements attain appropriately upscaled log-scale sensitivity. Furthermore, the sensitivity of the CSEM and MT methods was examined in other case histories, where the target reservoirs exhibited conductive properties, such as those encountered in enhanced oil recovery (EOR), geothermal, and lithium exploration applications. The same equipment specifications were utilized for CSEM and MT surveys across all case studies.

Barajas-Olalde, César↗
DOE OSTI2023

Comparison of quantum advantage experiments using random circuit sampling

Random circuit sampling, the task of sampling bit strings from a random unitary operator, has been implemented to demonstrate quantum advantage on the Sycamore quantum processor with 53 qubits and on the Zuchongzhi quantum processor with 56 and 61 qubits. Recently, it was claimed that classical computers using tensor network simulation could catch on to current noisy quantum processors for random circuit sampling. While the linear cross-entropy benchmark fidelity was used to certify all these claims, it may not capture statistical properties of outputs in detail. Here, we compare the bit strings sampled from classical computers using tensor network simulation by Pan et al. [F. Pan, K. Chen, and P. Zhang, Phys. Rev. Lett. 129, 090502 (2022)] and by Kalachev et al. [G. Kalachev, P. Panteleev, P. Zhou, and M.-H. Yung, arXiv:2112.15083] with the bit strings from the Sycamore quantum processor. It is shown that all of Kalachev et al.'s samples passed the NIST random number tests. The heat maps of bit strings show that Pan et al.'s and Kalachev et al.'s samples are quite different from the Sycamore or Zuchongzhi samples. The analysis with the Marchenko-Pastur distribution and the Wasssertein distances demonstrates that Kalachev et al.'s samples are statistically closer to the Sycamore samples than Pan et al.'s while the three datasets have similar values for the linear cross-entropy fidelity. In conclusion, our finding implies that further study is needed to certify or beat the claims of quantum advantage using random circuit sampling.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2024

Pretraining Billion-Scale Geospatial Foundational Models on Frontier

As AI workloads increase in scope, generalization capability becomes challenging for small task-specific models and their demand for large amounts of labeled training samples increases. On the contrary, Foundation Models (FMs) are trained with internet-scale unlabeled data via self-supervised learning and have been shown to adapt to various tasks with minimal fine-tuning. Although large FMs have demonstrated significant impact in natural language processing and computer vision, efforts toward FMs for geospatial applications have been restricted to smaller size models, as pretraining larger models requires very large computing resources equipped with state-of-the-art hardware accelerators. Current satellite constellations collect 100+TBs of data a day, resulting in images that are billions of pixels and multimodal in nature. Such geospatial data poses unique challenges opening up new opportunities to develop FMs. We investigate billion scale FMs and HPC training profiles for geospatial applications by pretraining on publicly available data. We studied from end-to-end the performance and impact in the solution by scaling the model size. Our larger 3B parameter size model achieves up to 30% improvement in top1 scene classification accuracy when comparing a 100M parameter model. Moreover, we detail performance experiments on the Frontier supercomputer, America's first exascale system, where we study different model and data parallel approaches using PyTorch's Fully Sharded Data Parallel library. Specifically, we study variants of the Vision Transformer architecture (ViT), conducting performance analysis for ViT models with size up to 15B parameters. By discussing throughput and performance bottlenecks under different parallelism configurations, we offer insights on how to leverage such leadership-class HPC resources when developing large models for geospatial imagery applications.

Tsaris, Aristeidis (aris)↗
DOE OSTI2020

Quantum state reduction: Generalized bipartitions from algebras of observables

Reduced density matrices are a powerful tool in the analysis of entanglement structure, approximate or coarse-grained dynamics, decoherence, and the emergence of classicality. It is straightforward to produce a reduced density matrix with the partial-trace map by tracing out part of the quantum state, but in many natural situations this reduction may not be achievable. We investigate the general problem of identifying how the quantum state is reduced given a restriction on the observables. For example, in an experimental setting, the set of observables that can actually be measured is usually modest (compared to the set of all possible observables) and their resolution is limited. In such situations, the appropriate state-reduction map can be defined via a generalized bipartition, which is associated with the structure of irreducible representations of the algebra generated by the restricted set of observables. One of our main technical results is a general, not inherently numeric, algorithm for finding irreducible representations of matrix algebras. In our work, we demonstrate the viability of this approach with two examples of limited-resolution observables. The definition of quantum state reductions can also be extended beyond algebras of observables. To accomplish this task we introduce a more flexible notion of bipartition, the partial bipartition, which describes coarse grainings preserving information about a limited set (not necessarily algebra) of observables. We describe a variational method to choose the coarse grainings most compatible with a specified Hamiltonian, which exhibit emergent classicality in the reduced state space. We apply this construction to the concrete example of the one-dimensional Ising model. Our results have relevance for quantum information, bulk reconstruction in holography, and quantum gravity.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2021

Draft ASME Boiler and Pressure Vessel Code Cases and Technical Bases for Use of Alloy 617 for Constructions of Nuclear Component Under Section III, Division 5

The American Society of Mechanical Engineers (ASME) Boiler and Pressure Vessel Code currently only allows five materials for use in construction of nuclear components for high temperature service. These are: 2.25Cr-1Mo and V-modified 9Cr-1Mo steels, Types 304 and 316 stainless steels and the high nickel Alloy 800H. Since 2005, the US high temperature gas-cooled reactor program has been characterizing elevated temperature mechanical properties of Alloy 617 as the leading candidate construction material for the intermediate heat exchanger. After analysis of these experimental results, along with historical data and additional results available through the Generation IV International Forum, Very High Temperature Reactor, Materials Program Management Board Materials Handbook, a draft ASME Code Case to allow nuclear construction with Alloy 617 for temperatures up to 1750°F (954°C) has been developed. This report contains the Code Case for Low Temperature Service Construction of Section III, Division 5, Subsection HB, Subpart A, Class A and Subsection HC, Subpart A, Class B components, which has been approved in Section II, Materials, and Section III, Rules for Construction of Nuclear Facility Components. Supporting technical justification for the low temperature Code Case is also included. This Code Case allows use of Alloy 617 up to 800°F (425°C). This report also contains an updated draft of a Section III, Division 5, Subsection HB, Subpart B, Class A Code Case for Alloy 617 to qualify it for use in construction of nuclear components up to 1750°F (954°C) for service life up to 100,000 hours. The draft contained in Appendix 4, subject to editorial revision and approval by the ASME Special Task Group on Alloy 617 Code Qualification, will be submitted for approval by letter ballot by the appropriate ASME Committees. The technical justification supporting the Code Case is presented in Appendix 5 of this report. This background document is part of the information package that will be submitted with the Code Case for ballot.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2021

Efficient Topology Design Algorithms for Power Grid Stability

The dynamic response of power grids to small disturbances influences their overall stability. This letter examines the effect of network topology on the linearized time-invariant dynamics of electric power systems. The proposed framework utilizes H 2 -norm based stability metrics to study the optimal placement of lines on existing networks as well as the topology design of new networks. The design task is first posed as an NP-hard mixed-integer nonlinear program (MINLP) that is exactly reformulated as a mixed-integer linear program (MILP) using McCormick linearization. To improve computation time, graph-theoretic properties are exploited to derive valid inequalities (cuts) and tighten bounds on the continuous optimization variables. Moreover, a cutting plane generation procedure is put forth that is able to interject the MILP solver and augment additional constraints to the problem on-the-fly. Finally, the efficacy of our approach in designing optimal grid topologies is demonstrated through numerical tests on the IEEE 39-bus network.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2023

Accelerating Random Forest Classification on GPU and FPGA

Random Forests (RFs) are a commonly used machine learning method for classification and regression tasks spanning a variety of application domains, including bioinformatics, business analytics, and software optimization. While prior work has focused primarily on improving performance of the training of RFs, many applications, such as malware identification, cancer prediction, and banking fraud detection, require fast RF classification. In this work, we accelerate RF classification on GPU and FPGA. In order to provide efficient support for large datasets, we propose a hierarchical memory layout suitable to the GPU/FPGA memory hierarchy. We design three RF classification code variants based on that layout, and we investigate GPU- and FPGA-specific considerations for these kernels. Our experimental evaluation, performed on an Nvidia Xp GPU and on a Xilinx Alveo U250 FPGA accelerator card using publicly available datasets on the scale of millions of samples and tens of features, covers various aspects. First, we evaluate the performance benefits of our hierarchical data structure over the standard compressed sparse row (CSR) format. Second, we compare our GPU implementation with cuML, a machine learning library targeting Nvidia GPUs. Third, we explore the performance/accuracy tradeoff resulting from the use of different tree depths in the RF. Finally, we perform a comparative performance analysis of our GPU and FPGA implementations. Our evaluation shows that for high accuracy targets, our GPU implementation yields 5-9x speedup over CSR, and up to a 2x speedup over cuML.

FPGA, Xilinx FPGA, GPU, Random Forest classificati↗
DOE OSTI2026

An Autonomous MCP Bridge to Rucio: Enhancing Data Management Accessibility for High Energy Physics

The Rucio Data Management System [1] is an important tool used by High Energy Physics experiments, including those at Fermi National Accelerator Laboratory, to store and manage exabyte-scale scientific datasets. Despite its central role in coordinating data across globally distributed storage sites, Rucio's command line interface (CLI) presents a steep learning curve, and makes it difficult for scientists to navigate through. To solve this issue, a containerized Model Context Protocol (MCP) [2] server was built that connects Large Language Models directly to Rucio, allowing AI agents to handle data tasks by using simple, natural language rather than memorized terminal commands. The core engineering focus of this project was moving the server away from slow terminal commands that require text parsing and replacing them with a native Python Client API toolset and a planned REST API framework. Moving to the Python API handles data operations directly in memory, which helps clear up formatting errors, provides the AI with clean, structured JSON data and speeds up tool execution. To prove that the system actually works, a benchmarking pipeline was also built with various questions to test the AI across four different model configurations. The questions included finding data scopes, tracking down specific datasets, and checking replication rules. Through benchmarking, early runs showed that with raw terminal text, the model would get confused and stuck, whereas switching to the Python API to feed the AI clean, structured data yielded massive improvement. By creating an intelligent and autonomous bridge to a storage network, this project shows how AI can be implemented in scientific data management, which ultimately helps scientists at Fermilab spend less time sorting through data and more time focusing on their experiments and analysis.

Akella, Kashyap [William Rainey Harper Coll.]↗
DOE OSTI2024

Exponential concentration in quantum kernel methods

Kernel methods in Quantum Machine Learning (QML) have recently gained significant attention as a potential candidate for achieving a quantum advantage in data analysis. Among other attractive properties, when training a kernel-based model one is guaranteed to find the optimal model’s parameters due to the convexity of the training landscape. However, this is based on the assumption that the quantum kernel can be efficiently obtained from quantum hardware. In this work we study the performance of quantum kernel models from the perspective of the resources needed to accurately estimate kernel values. We show that, under certain conditions, values of quantum kernels over different input data can be exponentially concentrated (in the number of qubits) towards some fixed value. Thus on training with a polynomial number of measurements, one ends up with a trivial model where the predictions on unseen inputs are independent of the input data. We identify four sources that can lead to concentration including expressivity of data embedding, global measurements, entanglement and noise. For each source, an associated concentration bound of quantum kernels is analytically derived. Lastly, we show that when dealing with classical data, training a parametrized data embedding with a kernel alignment method is also susceptible to exponential concentration. Our results are verified through numerical simulations for several QML tasks. Altogether, we provide guidelines indicating that certain features should be avoided to ensure the efficient evaluation of quantum kernels and so the performance of quantum kernel methods.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

nautilus : boosting Bayesian importance nested sampling with deep learning

ABSTRACT We introduce a novel approach to boost the efficiency of the importance nested sampling (INS) technique for Bayesian posterior and evidence estimation using deep learning. Unlike rejection-based sampling methods such as vanilla nested sampling (NS) or Markov chain Monte Carlo (MCMC) algorithms, importance sampling techniques can use all likelihood evaluations for posterior and evidence estimation. However, for efficient importance sampling, one needs proposal distributions that closely mimic the posterior distributions. We show how to combine INS with deep learning via neural network regression to accomplish this task. We also introduce nautilus, a reference open-source python implementation of this technique for Bayesian posterior and evidence estimation. We compare nautilus against popular NS and MCMC packages, including emcee, dynesty, ultranest, and pocomc, on a variety of challenging synthetic problems and real-world applications in exoplanet detection, galaxy SED fitting and cosmology. In all applications, the sampling efficiency of nautilus is substantially higher than that of all other samplers, often by more than an order of magnitude. Simultaneously, nautilus delivers highly accurate results and needs fewer likelihood evaluations than all other samplers tested. We also show that nautilus has good scaling with the dimensionality of the likelihood and is easily parallelizable to many CPUs.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Deep kernel methods learn better: from cards to process optimization

Abstract The ability of deep learning methods to perform classification and regression tasks relies heavily on their capacity to uncover manifolds in high-dimensional data spaces and project them into low-dimensional representation spaces. In this study, we investigate the structure and character of the manifolds generated by classical variational autoencoder (VAE) approaches and deep kernel learning (DKL). In the former case, the structure of the latent space is determined by the properties of the input data alone, while in the latter, the latent manifold forms as a result of an active learning process that balances the data distribution and target functionalities. We show that DKL with active learning can produce a more compact and smooth latent space which is more conducive to optimization compared to previously reported methods, such as the VAE. We demonstrate this behavior using a simple cards dataset and extend it to the optimization of domain-generated trajectories in physical systems. Our findings suggest that latent manifolds constructed through active learning have a more beneficial structure for optimization problems, especially in feature-rich target-poor scenarios that are common in domain sciences, such as materials synthesis, energy storage, and molecular discovery. The Jupyter Notebooks that encapsulate the complete analysis accompany the article.

97 MATHEMATICS AND COMPUTING↗